How a Resurrection Really Feels: A Bibliometric Analysis of the Vitality of Multi-Level Governance as a Concept1

Dion Curry Swansea University

James Callaghan 010, Singleton Park Swansea, UK, SA2 8PP +44.1792.295.628 (T) +44.7804.911.835 (M)

D.S.D.Curry@swansea.ac.uk Draft: Please do not cite without permission

1 The research leading to these results has received funding from the European Union’s Marie-Skłodowska

Curie EU H2020 action, grant agreement No. 661479 (Project MLGPIL), Socio-economic Sciences & Humanities.

Multi-level governance (MLG) is a contested concept within both the broader politics literature and specific sub-disciplines of political studies, where different conceptualisations and typologies of MLG both complement and compete with each other (for example, see Bache and Flinders, 2004; Stephenson, 2013). In addition, it is unclear where this academic debate can be placed in regard to practical discussions of new modes of governance, legitimacy, accountability, transparency and openness. This paper will explore the following research question: how has the concept of MLG developed in the social sciences literature over time? What are the academic and practical implications of the robustness of MLG as a concept? The main focus of the paper will be on examining the bibliometric breadth and depth of the concept in order to develop a conceptual map of MLG and how this differs across disciplines and sub-disciplines. Methodologically, the research will use bibliometric analysis to examine the idea of MLG across the social sciences, which provides a quantitative way of analysing over 1000 articles and almost 50,000 sources. This data can be used to analyse key research and sources used in conceptualising MLG, how the concept travels across disciplines and how research clusters on MLG exist statically and have developed over time. By identifying key articles, more fine-grained qualitative and quantitative analyses of the meaning of the concept of MLG can also be developed and theorised based on the breadth of the concept through the discipline and the depth of the concept in terms of cited references. The paper is highly relevant to understanding academic and practical applications of multi-level governance. First, it provides insight into how a concept develops, grows and spreads academically and across disciplines in both normative and analytical ways. This allows for an assessment of the robustness of this concept, with bibliometrics providing a quantitative approach to understanding the breadth and depth of literature on MLG. Second, it provides insight on new areas into which the concept is moving and how it can be theorised, thus highlighting where innovative approaches to conceptualisation may be found and how academic work on MLG can be linked to practical applications of the concept. The paper will first provide some background on the conception of multi-level governance, before discussing the methodological approach and implications of using bibliometrics to analyse a concept. Then, it will look at the conceptual breadth of MLG, as measured by the overal corpus of the literature on MLG. That will be followed by an analysis of the conceptual depth of MLG, as measured by the body of literature cited in MLG articles. The paper will then propose a modified framework for understanding and assessnig MLG before finishing with some conclusions about what the implications of this bibliometric analysis are for MLG as a concept and how MLG literature can move forward and continue to develop. Multi-Level Governance as a Concept Multi-level governance as a concept wades into an already crowded pool of related but distinct concepts of governance (such as network governance) over multiple levels (such as federalism or multi-level government). Multi-level governance was first coined by Gary Marks as a way of explaining structural funding processes in the European Union (1993). Originally, MLG aimed to provide an alternative and somewhat middle-ground theory to European integration that avoided both the state-centric nature of intergovernmentalism and the federalism (or federalism light) espoused by supranationalism and

neofunctionalism. MLG was conceived as a means of explaining how political actors at different political levels worked together to create and implement policies in areas with less clear-cut jurisdictional bounds than traditional policy. Two different types of MLG were identified. Type I MLG systems resemble federal-type structures, with non-intersecting, general purpose jurisdictions, clear spheres of authority and well-defined levels. In contrast, Type II MLG, a somewhat ‘newer’ governance form, is distinguished by overlapping, policy-focussed jurisdictions operating at shifting numbers of levels that are more flexible and inevitably messier than traditional federal-type multi-level structures. While MLG was initially used as a way of analysing EU-level processes, it has since broadened out to include regional (e.g. Bache and Andreou, 2011) and state-level analyses both within and outside the EU (e.g. Horak and Young, 2012) as well as bottom-up examinations of the roles of local governments (Grisel and van de Waart, 2011). This includes expansion into functional uses, where the concept was applied in new policy areas or country studies (Stephenson, 2013, p. 822) and even development of the concept as a way of identifying a normative ‘good’ form of governance (European Commission, 2001; Committee of the Regions, 2009). This creates the danger of conceptual stretching (Sartori, 1970) or the creation of a ‘container concept’ that tries to be everything to everyone (Van Geertsom, 2011, p. 169), but new research attempts to ‘travel’ the concept without reducing its precision or clarity. This has resulted in competing visions of how to study and explain multi-level governance (see, among others, Bache and Flinders, 2004; Conzelmann and Smith, 2008; Curry, 2015; Hooghe and Marks, 2003; Jordan, 2001; Kohler-Koch and Larat, 2009; Littoz-Monnet, 2010; Piattoni, 2010; Scharpf, 1997). The broadening and deepening of the concept in turn led to new attempts to categorise different modes of governance (see, for example, Grisel and van de Waart, 2011; Howlett, Rayner and Tollefson, 2009; Curry, 2015; March and Olsen, 1996; Offe, 2006; Scharpf, 1991; Treib et al., 2007; Weaver and Rockman, 1993). These all offer their own strengths and weaknesses, trying to strike a balance between comprehensiveness and elegance, general applicability and specific nuance. These refinements of the concept also raise new issues regarding MLG and its practical effects, including questions of scope (can MLG be applied outside of Europe, or to international relations?), academic rigour (is MLG a theory or just an organising framework?) and legitimacy and accountability (who is ultimately responsible when multiple elected and unelected actors at different levels are involved in crafting and delivering a policy?). Analytical Framework and Methodology Bibliometrics as a method of analysis grew out of library science and can be used to examine literature to reveal specific subjects, concepts or trends in large bodies of literature (Lawani, 1981; Hung, 2012; Bornmann, 2013) and has been used before in examining politics- and public administration-related concepts (Curry and Van de Walle, forthcoming; Vogel, 2013). Analyses looking at concepts can examine a full body of literature for

1. The quantity of literature on a subject;

2. The temporal span of this body of literature;

3. The types of literature;

4. The general importance of the literature, as measured by total citation counts;

5. The key titles for each field, as measured by citation counts for individual articles;

6. The origins and the spread of the literature (adapted from Lawani, 1981, p. 309).

Articles on multi-level governance were collected through Web of Science using the WoS Core Collection database. A topic search (which includes instances of multi-level governance in title, keyword or abstract) was conducted in all WoS journals, using both ‘multi-level governance’ and ‘multilevel governance’ variants. No lower year limit was set on the search, but the earliest articles on MLG dated to 1996 and an upper limit of 2015 was put on to avoid partial 2016 records. Once duplicate records were removed, this produced a body of 1046 articles from 1996 to 2015. The articles were then analysed using Bibexcel (available at http://homepage.univie.ac.at/juan.gorraiz/bibexcel/, Persson, Danell and Schneider, 2009) and VOSviewer. Analyses were conducted using WoS field codes, including title, author, journal, publication year, abstract, key words and author addresses. References cited within those texts, drawn from the WoS CD field code, were also analysed to determine conceptual depth. In the 1046 articles, a total of 43,076 citations were mined after being cleaned (by author and title). These citations were likewise analysed in Bibexcel to the full extent of the data, looking at cited authors, journals and years. The major limitations to this type of study are 1) the extent of the records kept by Web of Science, and 2) the extent of bibliographic information available on each article. WoS focuses mainly on articles, which can be seen as limiting. However, other sources – most notably books – are incorporated through the citation analysis, which covers any material cited and not only articles. Second, bibliographic information on the journal articles can be limited, especially as one explores further and further back through the years. For this reason, the analysis focuses on more recent years, and the large number of sources (1046 articles, over 43,000 citations) helps to limit the effects of any outlying or poor data. The article focuses on two analyses of the vitality of the concept – breadth and depth. Conceptual breadth looks at the body of literature on MLG itself to determine the spread of the concept through political science and potentially into other disciplines. This will be measured in three main ways. First, journal sources will be analysed to see where articles are published. Second, authors and countries of origin will be examined. Finally, title, abstract and keyword searchers were conducted.

Source of Analysis Sign of Conceptual Vitality

Journal sources

Authors and country

Titles, abstracts and keywords

Spread of journals

Spread of authors and author locations

Robust, discrete word co-occurrences

Conceptual depth examines the sources that articles on MLG cite. This allows for exploration of how authors on MLG engage with the wider literature. Here, three main measures can be used to assess the vitality of the literature. First, the extent of diffusion of references indicates whether a wide range of material is used to address the subject. Second, reference specialisation (i.e. lack of reliance on review publications) indicates a continued and updated approach to studying the concept. Finally, the usage of contemporary references shows that study has not stagnated by referring to older publications, and that authors continue to engage with new literature on the subject.

Source of Analysis Sign of Conceptual Vitality

Reference diffusion and co-citation

Specialisation of references

Spread of references

Lack of reliance on review literature

Reference age

and significant co-citation

Usage of contemporary references

Vitality of the body of literature can therefore be assessed in these two ways, looking at the literature itself and the references that that literature uses. Where possible, these data sources are compared to the body of literature on New Public Management. This comparison helps to establish reference points to another concept that is a similar age and with a similarly-sized body of literature (with 1069 articles written on NPM). Conceptual Breadth Articles and Journal Sources A total of 1046 articles on MLG were produced up to the end of 2015.

There was significant journal spread for these articles, with 361 unique journals publishing articles on multi-level governance. The most common journal was the Journal of European Public Policy with 48 articles on the subject, followed by Environment and Planning C – Government and Policy. 23 journals published more than 10 articles on the subject. Authors were also significantly spread, with 1679 authors over the 1045 articles. Only Ian Bache (14) wrote more than 10 articles on the subject, and a further 15 authors wrote more than five articles. In comparison to bibliometric analysis of other politics terms, the literature on MLG is relatively diffuse. The top ten journals publishing on ‘New Public Management’ produced 32.7% of overall articles. In comparison, the top ten journals publishing on MLG accounted for 22.5% of all articles. Given the highly specific terminology, the spread of journals can be seen as quite high. This diffusion is also evident in the Web of Science categorisation of the publications.

Top Web of Science Categories for MLG Articles (85 categories total)

0

20

40

60

80

100

120

140

160

19

96

19

97

19

98

19

99

20

00

20

01

20

02

20

03

20

04

20

05

20

06

20

07

20

08

20

09

20

10

20

11

20

12

20

13

20

14

20

15

Number of Articles Per Year

Number of Articles

1. Political Science (341) 2. Environmental Studies (321) 3. Public Administration (217) 4. Geography (116) 5. Economics (114) 6. Environmental Sciences (100) 7. International Relations (93) 8. Planning Development (88) 9. Urban Studies (78) 10. Law (51)

The literature was also remarkably fractured, given the specificity of the term. There were few ‘star’ authors with a significant number of publications on the topic, with the top author (Ian Bache) having 14 articles on the subject. There were 1679 authors in total, giving a significantly high author/article rate (1.61 authors per article, as compared to 1.38 authors per article for NPM). When looking at a weighted distribution of authorship,2 65 authors had written more than two occurrences on MLG. In contrast to other measures, this does indicate some weak core of MLG researchers, as the proportion of authors with multiple articles on the subject is higher than that found in NPM research. However, the central core, as measured by the top ten authors in NPM and their weighted proportion of overall papers, is similar to that shown in the NPM literature, with the top ten accounting for 4.7% of all papers, as compared to 4.5% for NPM. Author Diffusion The vast majority of the publications were journal articles (870), followed by proceedings papers (94), book reviews and reviews (94), editorial material (31) and book chapters (2). These were almost exclusively published in English (1014 of 1045 articles), with a few articles in German (13), Spanish (8), French (4) and Portuguese (2), as well as Norwegian, Lithuanian, Italian and Czech (1 each). Geographically, there was a significant spread of countries with authors writing about MLG, with 62 different countries producing articles on the topic. Despite its European origin, 3 of the top ten countries producing MLG literature are non-European, with the United States coming in second. The top 10 countries accounted for 93.6% of all articles. In comparison, the top ten countries for NPM literature accounted for only 77.1% of all articles, with 59 countries overall. This indicates that there was a more even spread across countries in the NPM literature, whereas it was more concentrated in fewer countries (with a long tail of countries with very few articles) for MLG. However, within the top ten, there was a more even spread between countries, with four countries having more than 100 articles on the topic, compared to only two countries for NPM.

Top Countries for MLG Articles

1. United Kingdom (260) 2. United States (154) 3. Germany (131) 4. Netherlands (107)

2 Single-authored papers are weighted as 1.0 to that author, two-authored papers are weighted as 0.5 to each

author, three-authored papers are weighted as 0.33 to each author, etc.

5. Canada (95) 6. Australia (52) 7. Italy (51) 8. Sweden (48) 9. Spain (45) 10. France (36)

The most cited article on MLG was Hooghe and Marks’ ‘Unraveling the Central State, but How? Types of Multi-Level Governance’ with 532 citations, followed by Adger et al.’s ‘Social-Ecological Resilience to Coastal Disasters’ and Folke et al.’s ‘Resilience and Sustainable Development: Building Adaptive Capacity in a World of Transformations’. Interestingly, three of the top ten articles were clearly outside the politics sphere of academic literature, and seven of the top ten articles dealt with environmental issues. Overall, 18 (1.7%) articles had more than 100 citations, which is slightly higher in comparison to the NPM literature (with 1.4% of articles having 100 or more citations). The corpus as a whole was similar to that of NPM literature, with 11,615 citations (without self-citations) in 9100 articles, an average of 12.77 citations per item and an h-index of 50. NPM articles, which had a similar number of articles, were cited 12,657 times in 9148 articles, an average of 13.21 citations per article and an h-index of 51. Title/Key Word/Abstract Usage Once multi-level governance and variations thereof were removed, a few notable words and phrases were prominent in the literature. ‘Climate’ and ‘Climate change’ were used frequently, with climate being the fifth most-used word in titles. Environmental issues were strong overall, with ‘Environmental’, ‘Water’, ‘Sustainability’ and ‘Sustainable’, ‘Biodiversity’ all placing highly as most commonly used words. There was still a strong focus on Europe in the literature, with ‘Europe’ being the second-most used word and ‘EU’ being the 11th most-used word. Structural terms such as ‘structures’ and ‘institutions’ were also prominent, as were related relational terms such as ‘relation’ and ‘network.

The papers covered a wide range of topics, but certain specific areas (either geographic or policy-related) could be discerned. In analysis of titles, ‘European’, ‘Europe’ and ‘EU’ (combined) were the second most-used terms (321 total occurrences), and ‘climate’ was the fifth most-used term (131 occurrences), with ‘Environmental’ (43) and ‘Water’ (41) also high on the list. Other signifiers, such as ‘Regional’ (and ‘regional’ at 125), ‘Local’ (57), ‘Urban’ (44) and ‘global’ (45) were commonly used in titles. ‘Climate change’ was also the third most-used key word (behind only ‘multi-level governance’ and ‘governance’), even higher than the European Union and Europeanization (fourth and fifth respectively, with 59 and 32 occurrences). However, there was not a significant overlap of key words, with only 35 being used in more than ten papers each. In terms of subject fields, Environmental Science & Ecology was the second most-used category as well (376 occurrences). Conceptual Depth Reference Diffusion and Co-Citation There were a total of 43,076 citations in the 1046 articles, slightly less than the number for NPM (over 47,000 citations). Hooghe and Marks’ 2003 American Political Science Review, ‘Unravelling the Central State, but How? Types of Multi-Level Governance’ was the most cited article, with 174 citations. Article citations reveal a few notable factors. First, perhaps unsurprisingly, Liesbet Hooghe is the most cited author, with 496 citations, followed by Gary Marks (437 citations), Harriet Bulkeley (346 citations), Elinor Ostrom (332 citations), Ian Bache (304 citations), Fritz Scharpf (230 citations), Michele Betsill (175 citations), Andrew Moravcsik (153 citations), Rod Rhodes (153 citations), Tanja Bӧrzel (147 citations), Bob Jessop (140 citations), Andrew Jordan (140 citations), Arthur Benz (139 citations), O. Young (132 citations), Michael Keating (131 citations), B Guy Peters (130 citations), Neil Brenner (120 citations), Barry Rabe (116 citations), Beate Kohler-Koch (114 citations), Fikret Berkes (108 citations). The European Commission (and its predecessor, the Commission of the European Communities) was also highly cited (at least 558 citations, although an accurate count is difficult given the number of ways in which this body is referenced), as was the OECD (173 citations). There were 231 articles cited at least 10 times, and 792 authors, as compared to 175 for NPM and 668 authors. There were 218 journals cited at least 10 times, compared to 143 for NPM. This indicates a reasonable spread of citations, with greater diffusion than was evident with NPM. While conceptual literature may get routinized over time, where certain sources – such as Hooghe and Marks – may become the ‘default citation’, this appears to be less the case with MLG than with NPM. Co-citation did not reveal any pairings between article citations that were unexpected (e.g. the most common pairs were between the most significant pieces of MLG literature). There did appear to be the genesis of two separate MLG bodies of literature – the traditional body of ‘governance’ literature and the newer environmental policy literature – but more analysis must be done to confirm this. In terms of subject depth, there was the noted move towards environmental policy, and expanding the scope to include non-European countries. In terms of general conceptual review articles, Hooghe and Marks’ ground-breaking work, and Bache and Flinders 2004 update, were the latest highly-cited pieces. As those are between 10-15 years old at this point in time, there appears to have been less development in further conceptualising MLG.

Specialisation of References While not a perfect proxy, the number of books that are heavily cited can be used to indicate a focus on review literature. 19 of the top 50 references were books, compared to over half of the top 50 in NPM, indicating a stronger journal presence for MLG. There was a similar spread of authors in the top 50 cited articles between the two terms, with 31 authors for NPM and 30 authors for MLG. However, there was more spread in the most heavily cited articles. While Hood and Pollitt dominated the most cited articles in NPM, there were at least six authors near the top of the most-cited list for MLG. While Hooghe and Marks are, to a certain extent, the ‘default’ authors of MLG, their citations are spread over multiple articles, and interspersed with significant numbers of citations to other authors as well. The move to MLG as an environmental concept has also increased the specialisation of references evident in the MLG literature. Usage of Contemporary References 32 of the top 50 references came after the year 2000, compared to only 15 for NPM, signifying a much newer literature for citations in MLG, and 65% of the top 100 articles dates from 2000 or later. The most recent articles in the top 100 most cited articles are six dating from 2010. The most cited year for MLG research was 2007, with 2830 citations. The top ten years for citations all came in the 2000s, with the earliest year being 2002 and the latest being 2011. These findings contrast sharply with the NPM literature, where references focused on much older literature, with a significant majority (over 65%) of the articles being written in the 1990s or earlier, and the latest article in a list of 100 most-cited articles dates from 2007. Over 22% of articles cited were from the previous 5 years,3 which is a common benchmark for the ‘newness’ of a concept (Price, 1970), a significantly greater proportion than the 8% of citations for NPM. These findings support bibliometric research and the concept of obsolescence (Line and Sandison, 1974; Sandison, 1987; Line 1993; Rowlands and Nicholas, 2007, pp. 226), which states that ‘literature of the past few years account for a large proportion of total citations’ in non-obsolete cases(Lawani, 1981, pp. 31; Meadows, 2005, pp. 91). Citation Years

3 As the cut-off date was 2015, the previous five years were 2010-2014.

The mean reference age is 10 years old, which puts it at significantly lower than that for NPM (16 years) and at the low end of average compared to (outdated) averages for economics (10.6), business (10.9) and sociology (12.5) (Glanzel and Schoepflin, 1999, pp. 41). MLG Vitality: A Broad and Deep Concept Overall, the bibliometric analysis points to MLG as a vital and current concept. In terms of breadth, the literature has spread into new areas, most notably environmental policy, and there is a wide spread of journals that write on MLG, even outside of the traditional political science and public policy channels. While the spread of countries with authors writing on MLG was somewhat more limited than the NPM literature, there was a larger number of countries producing significant numbers of articles on NPM, including many countries outside of Europe. This indicates that the concept has moved beyond its traditional EU trappings. In terms of conceptual depth, as measured by the works cited in the MLG literature, the concept also proved vital. There was a greater diffusion of references – both in terms of authors and journals – than there was in the NPM literature. There was a greater spread of authors being cited as well, and a heavier reliance on journal articles, rather than books, than NPM. Finally, references were significantly more current for MLG than NPM, with both a higher number of articles from the last five years being heavily cited and a much shorter average article age than NPM (10 years, as compared to nearly 16 years for NPM). Overall, this points to a vital concept that is still developing and moving in new directions. The quantitative analysis of MLG as a concept does point to some ways in which the concept can develop. While MLG is moving into new policy spaces, there is still a heavy reliance on original conceptualisations of MLG developed by Hooghe and Marks. In other words, MLG is being used in different ways, but the ways in which the term is understood have changed less and not been refined beyond the Hooghe and Marks’ typology. This points to more room to examine what MLG actually means, not just where it is evident. While the literature

0

500

1000

1500

2000

2500

3000

3500

4000

19

79

19

81

19

83

19

85

19

87

19

89

19

91

19

93

19

95

19

97

19

99

20

01

20

03

20

05

20

07

20

09

20

11

20

13

20

15

is vital enough to not require a prescriptive remedy for how to further develop it, conceptualisation of MLG is still somewhat underdeveloped and can be expanded upon. This paper now turns to a qualitative examination of MLG, along with a proposed refinement of the traditional MLG typology. Towards a Refined Conception of MLG The bibliometric analysis presented herein provides a broad-brush view of MLG as a concept, but a more qualitative analysis of the literature helps to provide nuance to our understanding of where MLG stands as a concept. The research presented now is based on research done on case studies in Canada (British Columbia) and Scotland on public sector service delivery, along with research done on political legitimacy in the UK and Wales. In addition, it incorporates research on intra-EU migration policy as part of the analysis of the IMAGINATION project led by Peter Scholten and Mark van Ostaijen.4 For reasons of brevity, this work is presented in short and mainly focused on the new conceptual framework that has been developed. A review of the literature reveals certain factors that go into shaping MLG, which can be broken roughly into structural, relational and policy factors. When governance plays out at multiple levels, attempts at coordination can face several factors that manifest themselves in structural, relational and policy ways. Complex institutional structures can lead to fragmentary policy-making. Relationally, hierarchy may make coordination difficult as actors compete for power. Finally, different actors may have competing, conflicting or shifting policy interests and goals (Taȿan-Kok and Vranken, 2011, pp. 16-17). These structural, relational and policy processes map onto Hooghe and Marks’ typology of MLG, but provide a more granular way of analysing the factors influencing MLG. The three categories of processes can be mutually reinforcing, contradictory or separate. Institutional structures and actor relations will have an impact on what policy options are open. The realities of policy-making and specific policy areas will, in turn, affect how actors work together and use institutional structures to develop policy. If these three factors are mutually reinforcing, structures that are supportive of MLG-type processes will develop. This will give actors more room to manoeuvre in shaping policy outcomes in a multi-level manner, actors may utilise structures in a way that supports multi-level solutions, and/or policies may lend themselves to solutions that make use of structures and relations in a multi-level way. If these processes are not mutually reinforcing, they can result in governance mismatch, which can take two forms. When these processes operate in contradictory fashion, disjointed governance can result, where actors, institutions and policies operate at cross purposes (Curry, 2015). Finally, when these processes operate separately, you find cases of decoupled governance, where there is little coordination between actors, institutions and processes (Scholten, 2013). Structures, relations and policy can be further broken down into 6 sub-categories.

4 http://project-imagination.net/

Table 1

Multi-Level Governance Processes in Intra-EU Migration Ty

pe

I M

ult

i-Le

vel G

ove

rnan

ce Structural Factors

Typ

e II

Mu

lti-

Leve

l Go

vern

ance

Rigid Institutional Structure

Multi-Purpose Jurisdictions

↔

↔

Flexible Institutional Structure

Policy-Specific Jurisdictions

Relational Factors

Hierarchical Relations

Formalised Relations

↔

↔

Heterarchical Relations

Informal Relations

Policy Factors

Discrete Policy Responses

Uniform Policy Issue

↔

↔

Holistic Policy Responses

Complex Policy Issue

Structural/institutional factors that affect the nature of multi-level governance can be grouped into two categories. First, there may be a rigid or flexible institutional structure in which actors operate. Rigid structures may be binding on the actors involved, which is often, but not always, tied to the threat (or not) of sanctions for inaction (or negative action). Conversely, these structures may take a flexible, non-binding, ‘soft law’ form that is more responsive to contextual and policy-specific factors. Second, the jurisdictions that make up the ‘multi-levelness’ of the policy area may be general purpose (normally geographical or cultural entities that exist across policy areas) or specific to that particular policy, an increasingly common configuration in EU settings. These two broad categories combine to create the institutional milieu in which governance exists, a spectrum where there can be clear lines of authority (in rigidly structure, general purpose jurisdictions), or one with indistinct lines of authority (flexible, policy specific approaches), and any variation in between. Relational factors are ones that affect how actors – whether governmental, quasi-governmental or non-governmental - interact with each other both within and between levels. Again, there are two broad categories of relational factors. First, relations may either be hierarchical with a clear chain of command, or be more heterarchical networks in which actors work more collaboratively. In areas where higher governmental levels have binding power, processes will be more likely to operate in a hierarchical manner, whereas heterarchical processes can originate where strong local levels exist, or where local levels assert themselves in areas left undeveloped by higher levels. In hierarchical, top down relationships, policy decisions are downloaded by decision makers at higher levels of government to be implemented by lower levels of government. Heterarchical relations allow for more two-way permeation of policy ideas, where processes may originate at lower levels and permeate upwards. Somewhere in between the two, lower levels of government may ‘upload’ information to decision-makers at higher levels of government (Güntner, 2011). While not a truly heterarchical process, as decision-making is not shared between all actors, uploading does allow for more input from actors at lower levels. Second, these relations may be highly formalised and standardised or may take a more fluid, informal or ad hoc approach. This stability or fluidity can operate in terms of context, policy response, actors involved or other factors. The spectrum of relations can vary from a highly inclusive

one marked by heterarchy and fluid, informal policy responses, to an exclusive one marked by hierarchy and stable, formal relations between actors. Finally, policy factors affecting MLG can be grouped into two categories. First, policy responses may be discrete and compartmentalised, where issues are treated separately with clear lines of demarcation. Conversely, a holistic approach treats each issue as part of a larger whole covering a broader policy area. Coordination is required for both approaches, but for different reasons. Discrete policy responses require coordination in order to clarify and designate the delegation of roles and responsibilities within specific issues, whereas holistic responses require coordination to ensure harmonisation across disparate issues. A lack of coordination can develop either between levels (for instance, with different governmental levels having different policies on an issue), within levels (where, for example, governmental and non-governmental actors may have different roles, policy goals and approaches) or over time. The level of coordination may also be tied to whether competence over that policy issue is exclusive or shared, and whether those creating the policy are the same as those implementing the policy. Finally, uncoordinated responses may develop when the policy issue is contested, either through new or unclear jurisdictional bounds or through attempts by actors to either take over or relinquish ownership and control of the policy. Second, the policy itself may be complex and cross-cutting or uniform, discrete and straightforward. Complex policies are ones that may deal with cross-cutting policy problems, have incomplete, contradictory or changing solutions, or have many possible options with no clear-cut ‘best’. Policy approaches are then found on a spectrum from chaotic (uncoordinated, contested, complex and non-state controlled), to controlled (coordinated, uncontested, straightforward, state-controlled policy). To relate this back to Hooghe and Marks’ initial typology, governance that mainly hews to the left side of Table 1 can be seen as mutually reinforcing in a way indicative of Type I MLG, whereas governance that fits on the right side is mutually reinforcing in a way closer to Type II MLG. Governance responses may in turn target structural, relational or policy factors. This can result in various governance outcomes depending on how complete the change is in the three identified factors. If there is no change in structural, relational or policy factors, the status quo will obviously ensue. If there is a change in only some of the factors, it may result in governance mismatch – either decoupled governance, where the levels operate more or less independently, or disjointed governance, where different levels may operate at cross purposes. A true multi-level governance response therefore must encompass complementary, mutually supporting shifts in all three factors. Conclusions This paper has shown that multi-level governance, almost 25 years on, remains a vital concept that has both breadth in the literature it produces, and depth in the literature it engages with. The paper gives quantitative evidence that MLG has expanded beyond its traditional EU confines to be used in both new national and policy contexts. Most notably, it has gained significant traction in non-European contexts and with non-European academics, and has a robust literature developing mainly in the field of environmental science and policy. While these new national and policy applications continue to thrive, less work has been seen in further developing what MLG actually means as a concept. This paper provides

a more refined conception of multi-level governance that further delineates types of MLG based on structural, relational and policy factors. The greater focus on specific policy areas, such as environmental policy, shows that MLG is growing on policy terms, while connections to concepts such as networks and institutions illustrate that structural and relational factors are also being taken into account. However, sources that can be seen as conceptual in nature (rather than based on specific policies or cases) were more limited and mainly date back to Hooghe and Marks’ and (to a somewhat lesser extent) Bache and Flinders’ initial work on the concept. While new case studies do increase the robustness and application of MLG, it remains important co consider the interplay between the structural, relational and policy considerations at play. Whether these multi-level governance processes are intended as an indicator of power relations, a tool of coordination, a tool of participation and influence, or something else will have an effect on what form the governance processes should take, and how successful they are in meeting those goals. These questions of what denotes MLG, what ‘successful’ MLG looks like and how this can be measured, are ones that the MLG literature could now look to tackle. There are some limitations to this type of analysis. While bibliometrics is useful in analysing an entire (large) corpus of literature in a way that cannot be easily accomplished using more qualitative methods, the approach lends itself more to a broader survey of ideas of a literature, rather than a deeper probe into the nuances of the literature. The next phase of this research project will be to incorporate a more qualitative reading of the literature, and the initial outcomes of this are presented in the second part of the paper that develops a new framework for MLG, although much of the qualitative analysis has been left out of this paper. The research can be developed in several directions in the future. First, the literature can be broken down into several component sub-sets, either based on discipline or focus (such as policy- or process-focused literature). In addition, this quantitative analysis will be combined with a qualitative analysis of articles that were identified as key or ‘important’ through the bibliometric measures. This will provide additional depth and richness to the understanding of what MLG itself means in the literature. Additionally, bibliometric analysis of more concepts will lead to a more robust understanding of where MLG literature fits in a broader political science context. This paper did compare MLG to New Public Management, but other concepts can be integrated in time. Finally, an analysis of the literature over time would provide a dynamic picture of how MLG has evolved over time and travelled within the discipline. However, this paper as it currently stands is intended to start a debate on how we can summarise the MLG research as a whole, and ways forward for how we conceptualise the term. Bibliography Bache, I. (2008). Europeanization and Multilevel Governance: Cohesion Policy in the European Union and Britain. Lanham: Rowman and Littlefield. Bache, I. and Andreou, G. (eds.). (2011). Cohesion Policy and Multi-Level Governance in South East Europe, Oxford: Routledge.

Bache, I. and Flinders, M. (2004). Multi-Level Governance. Oxford: Oxford University Press. Blom-Hansen, J. (2005). ‘Principals, agents, and the implementation of EU cohesion policy’, Journal of European Public Policy, 6(2), pp. 624-648. Bornmann, L. & Marx, W. (2013). The Proposal of a Broadening of Perspective in Evaluative Bibliometrics by Complementing the Times Cited with a Cited Reference Analysis. Journal of Informetrics, 7(1): 84-88. Committee of the Regions (2009). The White Paper on Multi-Level Governance. Available at http://cor.europa.eu/en/activities/governance/Documents/mlg-white-paper/0387_inside-en-last.pdf. Accessed 28 February 2016. Conzelmann, T. and Smith, R. (eds). (2008). Multi-Level Governance in the European Union: Taking Stock and Looking Ahead. Baden Baden: Nomos. Curry, D. (2015). Network Approaches to Multi-Level Governance: Structures, Relations and Understanding Power Between Levels. Basingstoke: Palgrave Macmillan. Curry, D. and Van de Walle, S. (2016, forthcoming, accepted). ‘A Bibliometrics Approach to Understanding Conceptual Breadth, Depth and Development: The Case of New Public Management’. Political Studies Review. Elmore, R. (1979-1980). ‘Backward Mapping: Implementation Research and Policy Decisions’. Political Sciences Quarterly, 94, pp. 601-616. ESPON. (2005). Project 3.2: Spatial Scenarios and Orientations in Relation to the ESDP and Cohesion Policy – Second Interim Report, Part 1. Available at http://www.espon.eu/export/sites/default/Documents/Projects/ESPON2006Projects/CoordinatingCrossThematicProjects/Scenarios/2.ir_3.2-full.pdf. Accessed 29 February 2016. European Commission (2001). European Governance: a White Paper. COM(2001) 428, Brussels: European Commission. European Council. (2016). ‘European Council meeting (18 and 19 February 2016) – Conclusions’. Available at http://docs.dpaq.de/10395-0216-euco-conclusions.pdf. Accessed 04 May 2016. Van Geertsom, J. (2011). ‘Postscript: The Need for a Common European Method of Multilevel Urban Governance’ in Grisel, M. and van de Waart, F. Multilevel Urban Governance or the Art of Working Together. Amsterdam: European Urban Knowledge Network, pp. 169-170. Grisel, M. and van de Waart, F. (2011). Multilevel Urban Governance or the Art of Working Together. Amsterdam: European Urban Knowledge Network.

Güntner, S. (2011). ‘Urban Development and European Forms of MLG’ in Grisel, M. and van de Waart, F. Multilevel Urban Governance or the Art of Working Together. Amsterdam: European Urban Knowledge Network, pp. 17-25. Héritier, A. and Lehmkuhl, D. (2008). ‘The Shadow of Hierarchy and New Modes of Governance’, Journal of Public Policy, 28(1): pp. 1-17. Hooghe, L. and Marks, G. (2003). ‘Unraveling the Central State, but How? Types of Multi-Level Governance’. American Political Science Review, 97, pp. 233-243. Horak, M. and Young, R. (Eds.) (2012). Sites of Governance: Multilevel Governance and Policy Making in Canada’s Big Cities. Montreal & Kingston: McGill-Queen’s University Press. Howlett, M., Rayner, J. and Tollefson, C. (2009). From Government to Governance in Forest Planning? Lessons from the Case of the British Great Bear Rainforest Initiative. Forest Policy and Economics, 11, 383-391. Hung, J.L. (2012). Trends of E-Learning Research from 2000 to 2008: Use of Text Mining and Bibliometrics, British Journal of Education Technology, 43(1): 5-16. Jack, S., Drakopoulou Dodd, S. and Anderson, A. (2008). ‘Change and the development of entrepreneurial networks over time: a processual perspective’, Entrepreneurship & Regional Development, 20(2), pp. 125-159. Jordan, A. (2001). ‘The European Union: an evolving system of multi-level governance…or government?’ Policy and Politics, 29(2), pp. 193-208. Kapucu, N. and Garayev, V. (2012). ‘Designing, managing, and sustaining functionally collaborative emergency management networks’, American Review of Public Administration, 43(3), pp. 312-330. Kelder, T. (2011). ‘Green Investment Programmes in Sweden: 1+1 Can be 3’ in Grisel, M. and van de Waart, F. Multilevel Urban Governance or the Art of Working Together. Amsterdam: European Urban Knowledge Network, pp. 150-156. Kohler-Koch, B. and Larat, F. (eds). (2009). European Multi-Level Governance: Contrasting Images in National Research. Cheltenham: Edward Elgar. Lawani, S.M. (1981). Bibliometrics: Its Theoretical Foundations, methods and Applications. Libri, 31(4): 294-315. Lipsky, M. (1980). Street-Level Bureaucracy: Dilemmas of the Individual in Public Services. Madison: University of Wisconsin Press. Littoz-Monnet, A. (2010). ‘Dynamic multi-level governance – bringing the study of multi-level interactions into the theorising of European integration’, available at http://eiop.or.at/eiop/pdf/2010-001.pdf. Accessed 28 February 2016.

Lowndes, V. and Skelcher, C. (2002). ‘The dynamics of multi-organizational partnerships: an analysis of changing modes of governance’, Public Administration, 76(2), pp. 313-333. March, J. and Olsen, J. (1996). Institutional Perspectives on Political Institutions. Governance, 9, 247-264. Marks, G. (1993). ‘Structural policy and multi-level governance in the EC’ in Cafruny, A. and Rosenthal, G. (eds.), The State of the European Community: The Maastricht Debate and Beyond. Boulder: Lynne Rienner, pp. 391-411. Mason, R. and Duval Smith, A. (2016). ‘Theresa May takes Brexit’s immigration message to eastern Europe’, The Guardian, 28 July, available at http://www.theguardian.com/politics/2016/jul/28/theresa-may-on-brexit-tour-of-eastern-europe. Accessed 01 August 2016. Offe, C. (2006). Political Institutions and Social Power: Conceptual Explorations. In I. Shapiro, S. Skowronek and D. Galvin, (Eds.), Rethinking Political Institutions: the Art of the State (p. 9-31). New York: New York University Press. Persson, O., R. Danell, J. Wiborg Schneider. 2009. How to use Bibexcel for various types of bibliometric analysis. In Celebrating scholarly communication studies: A Festschrift for Olle Persson at his 60th Birthday, ed. F. Åström, R. Danell, B. Larsen, J. Schneider, p 9–24. Leuven, Belgium: International Society for Scientometrics and Informetrics. Piattoni, S. (2010). The Theory of Multi-Level Governance: Conceptual, Empirical and Normative Challenges. Oxford: Oxford University Press. Sartori, G. (1970). Concept Misinformation in Comparative Politics. American Political Science Review, 64, 21. Scharpf, F. (1991). Political Institutions, Decision Styles, and Policy Choices. In R. Czada and A. Windhoff-Heritier, (Eds.), Political Choice: Institutions, Rules and the Limits of Rationality (p. 53-86).Frankfurt: Campus Verlag. Scharpf, F. (1997). ‘Introduction: the problem-solving capacity of multi-level governance’, Journal of European Public Policy, 4(4), pp. 520-538. Scholten, P. (2013). Agenda dynamics and the multi-level governance of intractable policy controversies: the case of migrant integration policies in the Netherlands. Policy Sciences, 46, 217-236. Stephenson, P. (2013). ‘Twenty Years of Multi-Level Governance: “Where Does It Come From? What is It? Where is It Going?”’ Journal of European Public Policy, 20, pp. 817-837. Taȿan-Kok, T. and Vranken, J. (2011). Handbook for Multilevel Urban Governance in Europe. Amsterdam: European Urban Knowledge Network.

Tosics, I. (2011a). ‘Multilevel government systems in urban areas’ in Grisels, M. and van de Waart, F. Multilevel Governance or the Art of Working Together. Amsterdam: European Urban Knowledge Network, pp. 26-35. Tosics, I. (2011b). ‘Urban development and urban policies in EU member states’ in Grisels, M. and van de Waart, F. Multilevel Governance or the Art of Working Together. Amsterdam: European Urban Knowledge Network, pp. 36-44. Treib, O., Bähr, H. and Falkner, G. (2007). ‘Modes of governance – towards a conceptual clarification.’ Journal of European Public Policy, 14, pp. 1-20. Vogel, R. (2013). What Happened to the Public Organization? A Bibliometric Analysis of Public Administration and Organization Studies. American Review of Public Administration, published online before print, January 15, 2013. van de Waart, F. (2011). ‘The Healthy Neighbourhood Experiment: a Multilevel Venture’ in Grisels, M. and van de Waart, F. Multilevel Governance or the Art of Working Together. Amsterdam: European Urban Knowledge Network, pp. 130-136. Weaver, R. and Rockman, B. (Eds.). (1993). Do Institutions Matter? Government Capabilities in the United States and Abroad. Washington: Brookings Institute.